Reuse of UV Mapping Space
polycounter lvl 3
Looking back at an old game I've modded, which does not have normal maps for its 3D models, I noticed some techniques for increasing texture resolution/detail.
1. The reuse of UV space, where very similar portions of geometry would have their UV islands ontop of eachother.
2. On symmetrical models, a center portion of a UV map would have a slight continuation past the mirrored portion allowing for asymmetrical texture features (e.g. an emblem, diagonal feature, etc...), where the texture would have to be carefully made to ensure the edge of that continued portion tiles back symmetrically.
These techniques in UV mapping now don't seem to apply properly in newer games because of how normal maps work.
Since the normal map texture requires the geometry to be at a specific position for the way lighting reacts to it to make sense, reuse of the same UV space for different geometry is impossible.
Not only this practice creates conflicting normals if the geometry is slightly different, but even if the geometry is identical but at different orientations.
Now that I can't do this I feel like I've lost any ability to optimize my use of UV space.
If I want to get better at this, what are some good guidelines to follow? For example, are there any possible mistakes in the topology of a mesh that could result in taking more UV space than necessary?
More than anything, if any experts here could just name some concepts for developing good topology and good UV mappings I would love to dig into those and practice their application.
1. The reuse of UV space, where very similar portions of geometry would have their UV islands ontop of eachother.
2. On symmetrical models, a center portion of a UV map would have a slight continuation past the mirrored portion allowing for asymmetrical texture features (e.g. an emblem, diagonal feature, etc...), where the texture would have to be carefully made to ensure the edge of that continued portion tiles back symmetrically.
These techniques in UV mapping now don't seem to apply properly in newer games because of how normal maps work.
Since the normal map texture requires the geometry to be at a specific position for the way lighting reacts to it to make sense, reuse of the same UV space for different geometry is impossible.
Not only this practice creates conflicting normals if the geometry is slightly different, but even if the geometry is identical but at different orientations.
Now that I can't do this I feel like I've lost any ability to optimize my use of UV space.
If I want to get better at this, what are some good guidelines to follow? For example, are there any possible mistakes in the topology of a mesh that could result in taking more UV space than necessary?
More than anything, if any experts here could just name some concepts for developing good topology and good UV mappings I would love to dig into those and practice their application.
Replies
http://wiki.polycount.com/wiki/Texture_Baking#UV_Coordinates
https://polycount.com/discussion/210340/use-of-trim-sheets-tiled-textures-or-baked-maps
That said, there are some optimization techniques that really aren't used anymore, such as mirroring tiled textures to eliminate the need for a cut along the mesh center (can't even find a good example of this anymore it's such an old technique). Another optimization technique that's gotten rare is where UVs are stretched to use every pixel on the map, which can save a lot of memory but at the cost of uneven pixel density (had a thread about it just a few months ago).